THE ANNALS OF HUMANITIES COMPUTING: THE INDEX THOMISTICUS
R. BUSX
Introduction
I entered the Jesuit order in 1933. I was then 20.
Later my superior asked me: "Would you like t o
become a professor?"
"In no way!" My wish was t o be a missionary to
take care of the poor.
"Good. You'll do it, all the same."
By 1941, when Italy entered the Second World
War, I had been assigned t o work towards a Ph.D. in
Thomistic philosophy at the Pontifical Gregorian
University in Rome. My research was aimed at
exploring the concept of presence according t o
Thomas Aquinas. At that time the Italian Navy was
interested in drafting me as a chaplain, an assignment
which would have greatly appealed to me. My superior, however, managed t o have someone else take the
post. Thus, up until the end of 1945, my principal
interest was focused on philosophy and philosophical
texts whlle I was surrounded by bombings, Germans,
partisans, poor food and disasters of all sorts.
According to the scholarly practices, I first
searched through tables and subject indexes for the
words of praesens and praesentia. I soon learned that
such words in Thomas Aquinas are peripheral: his
doctrine of presence is linked with the preposition in.
My next step was to write out by hand 10 000 3" X
5" cards, each containing a sentence with the word
in or a word connected with in. Grand games of solitaire foilowed. On January 28, 1946, I defended my
doctoral thesis, which was published in 1949: La Terrninologia Tomistica dell'lnteriorita: Saggi d i metodo
per una interpretazione della rnetaftsica della presenza
(Milano: Bocca).
While I was involved with this research, two major
considerations became evident. I realized first that a
Fr. Roberto Busa, S.J., CAAL-Aloisianum, 21013 G d a r a t e ,
Italy.
philological and lexicographical inquiry into the verbal system of an author has t o precede and prepare
for a doctrinal interpretation of his works. Each
writer expresses his conceptual system in and through
his verbal system, with the consequence that the
reader who masters this verbal system, using his own
conceptual system, has to get an insight into the
writer's conceptual system. The reader should not
simply attach t o the words he reads the significance
they have in his mind, but should try t o find out
what significance they had in the writer's mind.
Second, I realized that all functional or grammatical words (which in my mind are not 'empty' at all
but philosophically rich) manifest the deepest logic of
being which generates the basic structures of human
discourse. It is .this basic logic that allows the transfer
from what the words mean today t o what they meant
t o the writer.
In the works of every philosopher there are two
philosophies: the one which he consciously intends t o
express and the one he actually uses t o express it. The
structure of each sentence implies in itself some
philosophical assumptions and truths. In this light,
one can legitimately criticize a philosopher only when
these two philosophies are in contradiction.
In 1946 as a result of these preliminary conclusions, I started to think of an Index Thornisticus
(henceforth IT), i.e., a concordance of all the words
of Thomas Aquinas, including conjunctions, prepositions and pronouns, t o serve other scholars for analogous studies. This project has been considered favorably by such scholars as Aldo Ferrabino, Etienne Gilson, Werner Jaeger and the Jesuits Rink Arnou,
Charles Boyer, Paul Dezza, Ludwig Naber. It was
clear t o me, however, that t o process texts containing
more than ten million words, I had t o look for some
type of machinery. In 1949, accompanying an Italian
student, Giulio Crespi Ferrario from Busto Arsizio, I
visited approximately 25 American Universities from
84
R. Busa
/ The Index
coast t o coast, asking about any gadget that might
help in prod'ucing the type of concordance I had in
mind. Mr. H.J. Krould, Chief of the European Affairs
Division of the Library of Congress, provided me with
the answer in the person of Jerome Wiesner of M.I.T.,
who sent me t o IBM in New York City, where someone was assigned t o examine my project. I knew, the
day I was to meet Thomas J. Watson. Sr., that he had
on his desk a report which said that IBM machines
could never do what I wanted. I had seen in the
waiting room a small poster imprinted with the
words: "The difficult we do right away; the impossible takes a little longer," (IBM always loved slogans). I took it with me into Mr. Watson's office. Sitting in front of him and sensing the tremendous
power of his mind, I was inspired t o say: "It is not
right t o say 'no' before you have tried." I took out
the poster and showed him his own slogan. He agreed
that IBM would cooperate with my project until it
was completed "provided that you do not change
IBM into International Busa Machines." I had already
informed him that, because my superiors had given
me time, encouragement, their blessings and much
holy water, but unfortunately no money, I could
recompense IBM in any way except financially. That
was providential!
In addition Mr. Watson appointed Paul Tasman t o
assist me in the project. His broad mind realized
immediately the feasibility and values of the project,
as did Dorothy Tasman, his wife. His role in this project has been an essential one: without his contribution it could have failed many times. In the next days,
Cardinal Spellman informed Mr. Watson, through the
gracious efforts of Fr. Robert I. Gannon SJ, the
former president of Fordham University, that my
project was a valuable one, and that I had the proper
academic qualifications. In the three decades it took
t o complete the project, a time fdled with changes,
the support of my superiors, the management of IBM
and the Italian Financing Committees continued even
when it was not clear how long the project could
take.
When I returned to Italy, the IBM office in Milan
offered full assistance. I succeeded in producing single
word-cards from sentence-cards, progressively scanning the blanks in each column of the sentence-cards.
In 195 1 at the XVIII World Conference of Documentation held in Rome, my volume entitled S. 7'homae
Thornisticus
Aquinatis Hymnorum Ritualium: Varia Specimina
Concordantiarum: A First Example of a Word Index
Automatically Compiled and Printed by IBM
Punched Card Machines (Milano: Bocca 1951) was
exhibited. The many Italian IBMers who had collaborated were happy: G. Vuccino, Cl. Folpini, A. Cacciavillani (I regret that I might have skipped over
some names, but they know how grateful I am t o
them).
Although some say that I am the pioneer of the
computers in the humanities, such a title needs a
good deal of nuancing. A propos of this, Mr. Lee
Loevinger in the Minnesota Law Review, 3 3 (5)
April 1949), in an article on jurimetrics said: "Machines are now in existence which have so far imitated thought processes that they can solve differential equations. Why should not a machine be constructed to decide lawsuits?" (p. 471). And on the
stacks of the IBM library in New York City I had
spotted a book (whose title I have forgotten), which
was printed some time between 1920 and 1940: in it
someone mentioned that it was possible to make lists
of names by means of punched cards. Maybe others
too may claim that they have worked in this area
prior to me. Yet, isn't it true that all new ideas arise
out of a milieu when ripe, rather than from any one
individual? If I was not the one, then someone else
would have dealt with this type of initiative sooner or
later. To be the first bne having an idea is just chance.
If there is any merit, it is in cultivating the idea.
During the following years I experienced the wisdom
of another slogan, attributed t o an American, Thomas
Edison: Genius is 1% inspiration, 99% perspiration.
From punched cards to magnetic tapes
Obviously, I started with punched cards equipment, and as a consequence confronted two objectives. The first was t o establish a file of punched
cards, with one word on each card including reference
and typological codes. On the back of each card a
maximum of 12 lines, printed in the spaces between
the rows of holes, formed the context of the word
that was punched on the card. For this task I was
given an IBM 858 Cardatype, which was a kind of a
transitional link between unit record and data processing machines. Its input was made up of sentence-
R. Busa / The Index Thornisticus
ina
[ex
9M
vas
Ilaacver
to
he
a
,ee
5)
[ani:nInhe
ad
:h
it
:ts
rs
:a
se
le
se
3r
e.
a.
n
is
PC-
cards that were already punched and verified. It had
two outputs: on one side, a card, punched and interpreted, with reference and codes, for each single word
occurring in the sentence; on the other, a maximum
of 12 sentence-cards on a lithographic master plate.
This plate was then used (in an American Davidson
machme imported t o Gallarate and operated by a
Jesuit brother, Federico Masiero) t o print the context on the back of the corresponding word-cards. I
still remember how difficult it was t o calibrate the
lines between the punched holes, as the paper plate
stretched progressively during operation. I still have a
file of 800000 such cards. The second objective was
to find out a practical way for listing a concordance
using sentence-cards to be sorted and grouped for
each key-word.
One day I learned from the newspapers that an
Episcopalian minister, Rev. John W. Ellison, who was
preparing a concordance of the Revised Version of
the Bible (he published it in traditional ways in
1957), had used Remington magnetic tapes, which at
that time were not plastic but iron. 1 went t o shake
hands with him and said: "You are a great ally of
mine!" Immediately after I went t o IBM: "See what
Remington is doing?" Since that time the processing
of the IT has been done malnly by computers and
punched card equipment was used only peripherally.
A project of many facets
In Italy I worked in IBM offices for a few years. In
1954 I started my own punching and verifying
department; two years later I established my own
processing department, but employing large computers always in IBM premises. That year I started a
training school for keypunch operators. For all those
admitted, the requirement was that it was their first
job. After a month of testing, only one out of five was
accepted for a program of four semesters, eight hours
per day.. The success was excellent: industries wanted
t o hire them before they had finished the program.
Their training was in punching and verifying our
texts. To make the switch from the Latin t o the Hebrew and Cyrillic texts, only two weeks were needed,
and it was not even necessary t o attach these new
alphabets t o the keys of the puncher. In punching
85
these non-Roman alphabets, the process was less
speedy but with fewer errors. This school continued
until 1967, when I completed the punching of all my
texts. In that same year I moved my operations t o
Pisa, two years later t o Boulder, Colorado, and in
1971 t o Venice, wherever IBM provided computer
time.
In addition to the 10 600 000 words of the IT, I
processed five million more words, Italian, English,
German, Russian, ancient Greek and Hebrew, Aramaic and Nabataean, using Cyrillic, German Gothic,
Greek and Hebrew alphabets and going from right to
left in Hebrew. The subjects ranged from nuclear
physics and mathematics abstracts t o Qumran Scrolls
and works of Dante,Kant, and Goethe. The processing
of the Dead Sea Scrolls was publicized throughout
the world on the front pages of newspapers in the
spring of 1958.
The method for what I called the 'linguistic analysis' of natural texts had t o be tested. (I do not use
this expression in any of those specific and fluid
meanings which it has in some contemporary fields of
the philosophy of language.)
A parallel activity was the formation of the Promoting Committee for the IT, listing eminent Italian
scholars, and the establishment of a group of friends,
professionals and businessmen, into a Finance and
Administration Committee. (Among the founders,
Emilia and Nino Crespi of Busto Arsizio and Carlo
Pensotti of Legnano have already passed away.) Cardinal G.B. Montini accepted the presidency and later
Cardinal A. Luciani succeeded him. Was it mere accident that both these men who recognized the value of
wedding the old and the new were chosen to be
Popes?
Two more lines of my activity merged with my
daily teaching of philosophy at Gallarate College and
the production of the IT. One was the promotion of
similar projects in Europe. Up t o the present day I
have been active in around 6 0 conferences most of
which were international, from Russia t o California
and Brazil. Three have been organized and presided
over by me: in Tubingen University, November 2426, 1960 (Internationales Kolloquium uber maschinelle Methoden der literarischen Analyse und der
Lexikographie); in Pisa, March 27-29; l 9 6 8 (SCmimire intern. sur le Dictionnaire l a t h de Machine);
and in Venice, April 27-28, 1979 (Seminario di Lem-
86
JOS
RIC
THI
PHI
KA'
STE
MA1
I. G
WAI
THE
HU(
CH1
OK1
THC
STA
EM1
R01
AN1
HEE
BAR
PAL
DO1
DA\
FRE
ROE
DA\
WIL
HAP
HA1
WIN
DOI'
DOI'
DO I
CIA
Corn
scrip
currl
If nc
Subs
Ams
@N
rem'.
with
R. Busa / The Index Thomisticus
matizzazione Computerizzata). In 1967 Professor
Antonio Zampolli, up t o then my ,assistant, founded
in Pisa the laboratory for computational linguistics
which has made him internationally famous. In Tiibingen at the final session the audience wanted me to
accept the task of standardizing codes and methods. I
refused. "My Big Boss above wanted t o standardize
religion. See how He succeeded? Could I expect to be
more efficient than He?" Many centers, from Israel
to Czechoslovakia, Belgium, France, and Germany,
have been inspired by Gallarate Center.
Since 1949, I have visited America 35 times. First,
I needed to keep in touch with IBM technicians. My
second purpose was to exchange ideas with people
starting to make use of computers in scientific documentation, soon rebaptized as information retrieval
and now as information science. The late Prof. James
W. Perry, then at M.I.T. and later at the University of
Arizona, the author of many books, introduced me to
people, centers, publications and problems of the
field. Third, my activities became mixed with those
of machine translation, the gold rush which was dampened by the ALPAC report in 1966 that identified the
major obstacle to machine translation not as the
inadequacy of computer knowledge but rather as our
insufficient comprehension of natural language. At
that time I associated LCon Dostert's Georgetown
Project with the Euratom Center at Ispra, which is a
few miles from Gallarate.
Thirty years of work
In the meantime the operations at Gallarate
flowed along without interruption. From 1962 to
1967 we were a team of more than 6 0 full-time participants. Three main tasks were as necessary as they
were timeconsuming: pre-editing, lemmatizing, correcting. Various text-typologies had to be identified,
codified in the texts, and then properly punched in
the sentence-cards. For example, all sentences and
single words which the author quoted from other
authors had to be specified as such. I still react negatively to those who transfer from a text onto magnetic tape merely the unedited words and the punctuation marks, for in all texts there are features which
carry additional information, e.g., those from which a
reader understands that here the author quotes ver-
batim or summarizes a paragrah of another writer.
This information is lost when someone puts into
machine-readable form only the words and the punctuation. Our pre-editing demanded scanning all the
texts, word by word, at least twice. Associated operations were detecting text printing errors and fixing a
system to reference each word t o its location in the
text sequence. It was evident, furthermore, that in
text of 1 700 000 lines and 1 0 666 000 words, tables
and concordances should not represent only non-lemmatized graphic forms of words. Such unspecified
information in these quantities would be too bulky
and consequently useless. Therefore, a team of ten
priests worked with me for two full years to design a
Latin machine dictionary. Called the Lexicon Electronicum Latinum (LEL), it is a set of tables by which a
computer is able to lernmatize the words of Latin
texts. I defined lemrnatizing as two operations:
(1) grouping all the forms of an inflected word
(e.g., sum, es ... fui, fuisti ... essem, esses, fore,
futurus ...) under their 'lemma,' i.e., the title or
entry-word which' represents it in a dictionary (here,
the verb sum), and
(2) coding the morphological categories of each
form and lemma.
For this, we first punched, sequenced and numbered
the 9 0 000 lemmas in the Forcellini's Lexicon Totius
Latinitatis (Padua, 1940). Then, we alphabetically
sorted all the words of our texts and got a list of
130 000 graphically different forms of words occurring there. We next compared each of these forms
with the list of the lemmas. We established in which
lemma, either one or several, each of these forms
could belong. Finally we defined its morphological
categories. In this way, each word-form was examined
in isolation from any context; only *its generative
possibilities were considered. All problems and
aspects of ambiguity (I called it homography) had to
be faced and systematized. As the computer demands
full systematicity and absolute fullness, we had to
reorganize Latin grammar and the Latin lexicon. This
operation entailed tracing the borders between the
morphological and the syntactical; the distinction
between adjectives and nouns, for example, appeared
without doubt to be syntactical. It varies, in fact,
according to different contexts. Thus a lexicon
should provide this category as inducted statistically
from the uses of the word or as following semantic
R.Busa l T h e Index Thornisticus
er.
It0
1C-
the
raga
the
in
le S
mied
ky
:en
na
roha
tin
ns:
)rd
re,
or
re,
~ch
.ed
ius
lly
of
urms
ich
ms
cal
led
ive
nd
to
lds
to
his
.he
on
.ed
ct,
on
UY
tic
preclusions. At least in Latin, however, no morpheme
in the structure of a word ever differentiates an adjective from a noun.
The LEL now contains 150000 Latin forms fully
lemmatized and lemmatizing. When applied to any
other Latin text it will signal the forms which it does
not possess. The proceedings of the Pisa Seminar
(Revue [LiCge, 1969, pp. 1-176) record the features
which differentiate my 'morphological' Latin dictionary from the 'syntactical' one developed at LiCge
by Prof. L. Delatte, which lemrnatizes every occurrence of a word in its contest. The ten fields of my
morphological codes plus the four fields of homography codes of my LEL could be defective, but if so,
only in the line of too much and not of too little.
I had completed the keypunching of all my texts
before the opportunity of correcting texts on tape at
a video-terminal existed. Less than 20% of the time
was spent on the first punching and more than 80%
on cleaning the input. Verifying all the text-cards on
a verifier, we corrected the errors and then listed the
corrected cards. In teams of two, we checked the list
by reading it against the text. Then we repeated the
process by listing, checking and correcting everything
again. Even after these three checks, however, we still
discovered 1600 punching errors and one full line lost
because of an homoteleuton.
The ratio of human work t o machine time was
more than 100 : 1. Computer hours were less than
10000 while man hours were much more than one
million. In fact we had to scan our texts word by word
with human eyes and fingers at least 9 times: twice
for pre-editing; once for punching; once for verifying;
twice (two people each) for checking; twice for lemmatizing and sorting the homographs; and once for
final arrangements and checkings. All this would be
about equal to scanning 95 000000 separate words.
That means that in 25 years we processed an average
of 2200 words per working hour or 4 lines of text per
minute. I imagine that a similar quantity of text
could be processed today in ten or even six years,
using video terminals and optical character recognition. But then I had to solve problems which no
longer exist today. Without assistance and in addition
to fmding financial support, I had to develop and test
a method which had no predecessor and had to use a
technology which developed progressively.
87
Retrospective criticism
I was never trapped in a major cul de sac or distracted by a major U-turn. Nevertheless I realize
already three defects in my procedures. First, as we
processed the reference as a field in each line- and
word-record, it would have been more practical to
have the reference as a record in itself and to develop
it automatically into each word reference only after
the final text corrections: we could have avoided
needing to correct the referencing of all the following
words when adding or deleting a word. Then, in the
Concordantia Prima, we sorted the key-words according to the contiguous following word only when it
was one of those having a grammatical or connective
function. It would have been more useful, in sorting
each key-word, to attach to it any type of following
word (obviously with no major punctuation
between), as D.W. Packard did in his Livy concordance. Finally, in the Concordantia Altera all trinomiums having the same words were sorted first by
speech typologies and then by their punctuation
marks. It would have been better to do vice versa,
i.e., to sort them first by punctuation and then by
speech typology.
Were I to begin a similar project today, however,
I would process it the same way. After 3 0 years of
research I am still convinced that:
pre-editing and lemmatizing are necessary in processing large texts;
an important scientific role is played by processing of function and high-frequency words (pronouns, et, non, sum, etc.); this was almost never done
previously because it is infeasible manually, but it is
practical using a computer;
adding summary tables with quantitative information to concordances provides so valid a research tool
that I cannot see any good reason for omitting it.
I was able to complete my IT 33 years after the conception of the project and 3 0 years after my first
meeting with IBM. At that time it was the first undertaking in computer linguistics. Even today it is the
only published work of its kind with such dimensions
and such characteristics. I feel like a tight-rope walker
who has reached the other end. It seems to me like
Providence. Since man is child of God and technology
is child of man, I think that God regards technology
88
JOS
RIC
THI
PHI
KA'
STE
MA1
I. G
WAI
THE
HU(
CHd
OK1
THC
STA
EM1
R01
AN1
HED
BAR
PAL
DO1
D A\
FRE
ROE
D A\
WIL
HAE
HAI
WIN
DOP
DOP
DOI
GIA
Subs
Ams
@N
retri
with
Sub!
sum;
cop:
191;
R. Busa / The Index Thornisticus
the way a grandfather regards his grandchild. And for
me personally it is satisfying to realize that I have
taken seriously my service to linguistic research.
Anyone comparing the typographical quality of
the IT with offset computer printouts will understand
why I am now happy that I did not complete the project before photocomposition was available. Before
publishing the IT in book form, we'debated publishing it on microfiche or simply keeping it available
in a data bank. Since half of the 500 printed copies
were sold in two years, it is clear that we made the
right decision. Research with books or on a computer
terminal should not yet be considered alternative but
rather complementary. In most cases only printed
volumes can provide the information necessary for
computer-aided research, and very often one can
locate in a few seconds in the printed volumes information whch would require the scanning of 10 million records if it were to be located on tapes.
The computerization of language in its present state:
a personal reflection
In sketching the status artis in the electronic processing of language, spoken language and written
language must first be distinguished. By processing
spoken language I mean speech synthesis and speech
analysis, which is much more than recording, transmitting, reproducing the human voice. The techniques by which the computer has a human voice as
its output have already surpassed the laboratory
phase. In speech analysis a computer has a human
voice as its input, as in dictation to a Typewriter. These
techniques-are still caught in slow and laborious laboratory research,
Written language has to be divided into handwritten and printed texts. Here, too, input and output must be distinguished. Post offices are extremely
interested in a computer that would read any kind of
handwriting. It seems that, altogether, these techniques have not gotten beyond the research stage. In
any case, an electronic deciphering of ancient manuscripts would be science fiction.
The electronic reading of typed texts, on the other
hand, whenever all letters are inscribed in identical
spaces, has been in practical use for many years
already. But only since a few years ago has the tech-
nique. been perfected for reading a printed text where
the letter spaces are unequal. As for electronic printing, the 70000 photocomposed pages of my IT bear
witness that today it offers no fewer output possibilities and qualities than manual or hot typesetting.
Concerning the status artis of the procedures
themselves, I need to distinguish two kinds of
'words': on one hand, all digits, any other system of
symbols (e.g., musical notes or codings for archeological findings) and, surprisingly enough, proper
names; on the other, those words which we commonly call words. The processing of digits and symbols by computer in mathematics, sciences, documentation and all business has reached gigantic proportions. Today's technological explosion is based upon
it. The computer processing of those words which we
commonly call words, however, has not yet developed beyond the first meager steps. We must not hide
the fact that the compiling of tables, indices and
concordances of individual words, including their correlations, though so necessary for philology and
lexicography, is conceptually a somewhat feeble
result, still very far from practical applications. We
shall have an information industry in the full sense of
the term only when we have computer programs performing indexing and abstracting operations.
The major obstacle lies in the 'semanticity' of
'words,' which is deeply different from that of numbers and symbols. There is a multiple diversity; only
in words, for instance, is metaphor possible. Furthermore, we do not speak in words but in sentences. A
sentence has a global meaning which is not the pure
sum of the values of its single components. The heart
of this problem is whether or not we are able to formalize the global meaning of sentences with something less than the whole sentence itself; in other
words, whether or not we can succeed in identifying
in each sentence something which can be taken as
characteristic of its global meaning. I am not sure
whether artificial intelligence will eventually solve
this problem.
Computerization provides humanities with new qualitative dimensions
The use of computers in linguistic analysis is first
of all very different from scientific computation: the
R. Busa
i
/ The Index
latter has comparatively little input. little output,
huge processing; the former has huge input, huge output, comparatively little processing. One could say
that our job is similar to computerizing bank
accounts, but there is still a tremendous difference:
bank processings are repetitive functions, while in linguistic research, precisely because it is research, most
of the programs are used only once, for the phase for
which they have been designed; the following phase
will demand a new and different program. In other
fields, the computer is used to give those who have to
make decisions about events a summary of a flow of
those events as they take place. In these cases the
computer should be as close as possible to be contemporaneous with the events. In computerizing linguistic analysis there is obviously no such urgency.
In this field one should not use the computer primarily for speeding up the operation, nor for minimizing the work of the researchers. It would not be
reasonable to use the computer just to obtain the
same results as before, having the same qualities as
before, but more rapidly and with less human effort.
Imagine a research project which without the computer would require a man to work one year: in my
opinion it would not be the optimal use of the computer to complete the same research in a month. But
the optimal and specific use of the computer would
be for two years on a research project one thousand
times larger and one hundred times more profound,
aiming at results which would be unobtainable without it. During these two years the researcher's work
will not be diminished but rather increased, yet concentrated in those higher levels of analytic and creative functions whlch are the prerogative of the human
mind. To repeat: the use of computers in the humanities has as its principal aim the enhancement of the
quality, depth and extension of research and not
merely the lessening of human effort and time.
In fact, the computer has even improved the quality of methods in philological analysis, because its
brute physical rigidity demands full accuracy, full
completeness, full systematicity. Using computers I
had to realize that our previous knowledge of human
language was too often incomplete and anyway not
sufficient for a computer program. Using computers
will therefore lead us to a more profound and systematic knowledge of human expression; in principle, it
can help us to be more humanistic than before.
Thornisticus
There is another reason for saying this: insofar as
linguistic research attacks the semantic content of the
words, sampling is no longer sufficient. Whereas it can
be a valid method, in phonetics, for example, it
cannot be in lexicography. Linguistic research must
tend to be based on full inventories of words at least
of large corpuses. Semantics is the description of the
kingdom of the creative freedom of spirit: in man it is
his soul which talks. Consequently scientific description of how we talk can be nothing but probabilistic.
Today's academic life seems t o be more in favor of
many short-term research projects which need to be
published quickly, rather than of projects requiring
teams of co-workers collaborating for decades. But,
going back t o what I have just said, to put into practice the electronic processing of human sentences as
such, much more induction is needed. The magnificent store of mathematical methods we have today
has to be based on linguistic censuses of natural texts
of millions of words. Sometimes a splendid amount
of mathematics is applied to too small a base of linguistic data. It would be much better to build up
results one centimetre at a time on a base one kilometre wide, than to build up a kilometre of research
on a one-centimetre base.
The human factor in the computerization process
A few final notes on the type of human work
required. In terms of time, the most demanding phase
has been the preparation of the input. Only a few
organizing decisions have been conceptually complex
and difficult. Most of the human operations were
simple but had to be repeated over and over again
tens of thousands of times. Particular attention was
given to the necessity of being consistently inductive
and analytical; that is to say, ready to recognize only
those categories effectively emerging from the data.
In other words, we endeavored to avoid forcing the
data into preformulated or imaginary a priori categories. The computer has been a great help in this
process, as it allowed us to check the validity of our
categories on the full inventories of all involved data.
Other aspects of human commitment required in
this project were those Comnon to any teamwork,
yet two were given special care. The first was utmost
attention to every minimal detail. A dogma was that
90
JOS
RIC
THI
PHI
KA'
STE
MA1
I. G
WAI
THI
HU(
CH)
OK1
TH(
STA
EM1
R01
AN'I
HE1
BAN
PAL
DO I
DA\
FRE
ROE
DA\
WIL
HA?
HAF
WIN
DOP
DOP
DO 1
CIA
scrip
CUIIi
If nc
Sub!
Ams
re trz.
with
Sub1
sum,
COP.
191;
R. Busa
/ The Index
no one should allow himself, or allow any other person, to overlook an error or a defect or a doubt on
the assumption that it was a small one and seemingly
of no importance. No one can afford to ignore even
a single loose screw in a machine or the entire works
may fail. The second was the need to armor oneself
with inexhaustible patience and perseverance to cope
Thornisticus
with snags, accidents, machine failures, errors and ur
foreseen events, which have rendered linguistic ana
ysis very similar to an obstacle race.
That is the reason why the use of computers i
linguistics demands a lot of dedication and har
work. Without them, computers would only produc
'in real time' monuments of waste.